Retinoic acid receptor γ activation promotes differentiation of human induced pluripotent stem cells into esophageal epithelium

Background The esophagus is known to be derived from the foregut. However, the mechanisms regulating this process remain unclear. In particular, the details of the human esophagus itself have been poorly researched. In this decade, studies using human induced pluripotent stem cells (hiPSCs) have proven powerful tools for clarifying the developmental biology of various human organs. Several studies using hiPSCs have demonstrated that retinoic acid (RA) signaling promotes the differentiation of foregut into tissues such as lung and pancreas. However, the effect of RA signaling on the differentiation of foregut into esophagus remains unclear. Methods We established a novel stepwise protocol with transwell culture and an air–liquid interface system for esophageal epithelial cell (EEC) differentiation from hiPSCs. We then evaluated the effect of all-trans retinoic acid (ATRA), which is a retinoic acid receptor (RAR)α, RARβ and RARγ agonist, on the differentiation from the hiPSC-derived foregut. Finally, to identify which RAR subtype was involved in the differentiation, we used synthetic agonists and antagonists of RARα and RARγ, which are known to be expressed in esophagus. Results We successfully generated stratified layers of cells expressing EEC marker genes that were positive for lugol staining. The enhancing effect of ATRA on EEC differentiation was clearly demonstrated with quantitative reverse transcription polymerase chain reaction, immunohistology, lugol-staining and RNA sequencing analyses. RARγ agonist and antagonist enhanced and suppressed EEC differentiation, respectively. RARα agonist had no effect on the differentiation. Conclusion We revealed that RARγ activation promotes the differentiation of hiPSCs-derived foregut into EECs. Electronic supplementary material The online version of this article (10.1007/s00535-020-01695-7) contains supplementary material, which is available to authorized users.


Introduction
The esophagus is known to be derived from the anterior region of the foregut via tracheal-esophageal separation in embryogenesis. Previous studies using mouse models have identified several important signals in the development of esophageal epithelium [1,2]. However, the molecular mechanisms regulating this process remain to be fully clarified. In particular, relatively little information is available concerning the developmental process of human esophageal epithelial cells (EECs).
Human tissue organoids derived from human pluripotent stem cells have proven powerful tools for research on the developmental biology of various human organs [3]. For example, several reports using *human-induced pluripotent stem cells (hiPSCs) have identified signals and transcriptional factors that play essential roles in the development of foregut-derived organs, such as the liver, lung and stomach [4][5][6]. However, there have been only two reports concerning the generation of human esophagus from iPSCs [7,8], and neither resolved molecular mechanisms underlying the differentiation protocols employed.
All-trans retinoic acid, which binds to RARa, RARb and RARc as a pan RAR agonist, has shown to promote differentiation of various endodermal tissues, including lung, pancreas and bladder [5,9,10], and tissues consisting of squamous cells, such as skin and cornea [11,12], in several studies using human cells. In mouse models, retinoid acid signaling enhanced the terminal differentiation and proliferation of esophageal progenitor cells [13,14], but the effects of retinoid acid signaling on the earlier developmental processes of esophagus have never been reported. Regarding the process of differentiation from human foregut into esophageal epithelium, the effect of retinoic acid signaling remains unclear.
In this study, we generated human esophageal epithelium from hiPSCs using a novel stepwise differentiation protocol and evaluated the effect of retinoic acid signaling on the processes. We found that RARc activation promotes the differentiation of human anterior foregut into EECs.

Methods iPSC culture
We used the iPSC line FF-PB-3AB4, which was established from a healthy donor's peripheral blood mononuclear cells (PBMCs) [15]. The institutional review board of the Kobe University Graduate School of Medicine approved this study (No. 1722), and informed consent was obtained from the donor.

RA receptor-specific agonists and antagonists
To assesses which RA receptor subtype plays an essential role in esophageal differentiation, 1 lM of the RARa agonist AM580 (Cayman Chemical, Ann Arbor, MI, USD), 2 lM of the RARb agonist CD2314 (R&D Systems) and 0.5 lM of the RARc agonist BMS961 (R&D Systems) were added from Day 6 to Day 21 ( Fig. 4)  Transwell culture with air-liquid interface (ALI) For the in vitro 3D reconstruction, 1.8 9 10 5 iPSCs were seeded onto transwells (0.4-lm pore size; Corning, Corning, NY, USA) pre-coated with iMatrix-511 silk (0.5 lg/ cm 2 ). The protocol of differentiation was the same as that of normal culture. The ALI procedure was started on Day  Immunocytochemistry Cells were fixed with 10% formalin for 10 min at room temperature. After being washed with PBS, the cells were treated with 0.2% Triton X-100 in PBS and blocked with 5% donkey serum and 1% BSA in PBS for 45 min at room temperature. The cells were incubated with primary antibodies at 4°C overnight and then stained with secondary antibodies. The primary antibodies were rabbit anti-SOX2 (ab97959, dilution 1:200; abcam, Cambridge, UK) and goat anti-p63 (BAF1916, dilution 1:200; R&D Systems). All of the secondary antibodies (Alexa Fluor 488-conjugated anti-rabbit IgG and Alexa Fluor 594-conjugated antigoat IgG) were obtained from Life Technologies. Hoechst 33,342 (WAKO) was used for nuclear staining.

Frozen section samples
The cultured organoids were fixed with 10% formalin and then embedded in Tissue-Tek O.C.T Compound (Sakura Finetek Japan, Tokyo, Japan) and frozen at -80°C. The frozen samples were sectioned at 8 lm on a cryostat.

Immunohistochemical analyses of the organoids
For immunohistofluorescence, the sections were treated with 0.2% Triton X-100 in PBS and blocked with 5% donkey serum and 1% BSA in PBS for 45 min at room temperature. Primary antibodies were incubated overnight at 4°C. Slides were washed in PBS and incubated with secondary antibody for 1 h at room temperature. The primary antibodies were rabbit anti-SOX2 (ab97959; abcam),

Lugol staining
After being washed with PBS, the cells were treated with lugol solution (1.2% potassium iodine and 0.6% iodine in water) and then washed again with PBS. Lugol solution was kindly provided by Kobe University Hospital (Kobe, Japan).

RNA sequencing
Total RNA was isolated using Trizol (Life Technologies) and treated with the TURBO DNA-free kit (Thermo Fisher Scientific), as described above. The RNA was sent to Macrogen (Seoul, South Korea, https://www.macrogen. com) for library preparation and paired-end RNA sequencing on the Illumina Novaseq6000 platform. Raw sequence files (fastq) were aligned to the human Extraction of esophagus-specific genes and Z score transformation We defined tissue-specific markers as those with a [ fivefold higher average Fragments per kilobase of exon per Million mapped fragments (FPKM) level in esophagus tissue than in all other tissues [18]. After normalizing with TPM, we performed log base two transformation. The average Z-scores of tissue-specific genes were then calculated using Excel 2016 (Microsoft, Redmond, WA, USA), and a heat map was created using Excel 2016 (Microsoft).

Statistical analyses
All data were analyzed using the JMP Ò 10 software program (SAS Institute Inc., Cary, NC, USA). Differences in the mean values between two groups were analyzed using the paired t test, and three groups were analyzed using an analysis of variance Tukey test. The differences were considered to be statistically significant for P values \ 0.05 (*).

Induction of EECs from hiPSCs
Based on previously reported FG endoderm induction protocols [4-6, 10, 19-22], we developed an original stepwise differentiation method from hiPSCs into EECs via DE, FG and dAFG. First, hiPSCs were induced to differentiate into DE by treatment with the WNT activator CHIR99021 (hereafter CHIR) and Activin A for 3 days. We then differentiated DE into FG with a combination of LDN193189 (hereafter LDN) and SB431542 with CHIR. We confirmed the obvious expression of the FG marker genes SOX2 and FOXA2 by RT-PCR and immunocytochemistry (S- Fig. 1a and b). Since previous reports have shown that efficient FG induction requires an appropriate concentration of CHIR during DE induction [15,19], we examined the expression of the FG marker SOX2 and mid/hindgut marker CDX2 at Day 6 when cultured with 1 or 3 lM of CHIR during DE induction (S- Fig. 1c) [23]. Quantitative PCR showed that 1 lM CHIR treatment induced a significantly higher expression of SOX2 and lower expression of CDX2 than 3 lM CHIR treatment (S- Fig. 1d). Based on these results, we used 1 lM of CHIR for DE induction in our protocol.
To induce differentiation of dAFG cells into esophageal lineages, we tested three conditions after Day 13 (Fig. 1d). The first condition involved continued culture under the same conditions as before Day 13 (CFKBE ? ATRA). Under the second condition, we used Sup-KGM (see the ''Materials and Methods ''section), as we previously succeeded in generating bladder urothelium, which shares stratified epithelial structures with esophageal epithelium, using Sup-KGM [15]. Under the third condition, we omitted lung specification reagents (CHIR, FGF10 and KGF) [20] from the CFKBE ? ATRA cocktail and added Sup-KGM. After 8 days' culture, the expression of esophagus marker genes (SOX2, p63 and PAX9) were detected under all three conditions. However, the expression of the mature esophageal epithelial markers CK5 and CK13 [21] was detected only in the cells generated under the third condition (Sup-KGM ? BE ? ATRA) (Fig. 1e).

Expression of esophageal marker genes in the differentiated cells with ALI culture
A stratified structure is an important property of esophageal epithelium. We, therefore, employed ALI culture on a membrane (transwell culture system) as ALI culture enable us to assess vertical sections of the epithelium [25,26] and has been reported to induce stratified structure formation of primary EEC culture [27]. Our differentiation protocol and the marker genes at each step are shown in Fig. 2a. On Day 24, we performed immunohistochemistry for the derivatives. We deduced that, like derivatives of hiPSCs in previous studies [7,8], the derivatives in the present study would exhibit fetal histology but not adult histology, which would probably require more long-term differentiation culture. We,, therefore, used esophagus tissues of fetal mice (E17.5) as positive controls, as we were unable to obtain human fetal esophagus, which was the preferable positive control. Mouse esophagus is known to have keratinized layers, while human esophageal epithelium is not keratinized, and cells retain their nucleus [21]. Our HE staining findings confirmed this species difference in keratinization (S- Fig.3).
All of the esophageal marker genes we tested (SOX2, p63, CK5, E-cadherin, CK13, PAX9, Claudin-4 and Involucrin) were obviously expressed in the differentiated cells from hiPSCs ( Fig. 2b and S-Fig. 3, upper panels), similar to the esophagus of mice (E17.5) (Fig. 2b and  S-Fig. 3, lower panels). The findings of HE staining (S- Fig. 3) and the obvious expression of cell-cell adhesion molecules (E-cadherin and Claudin-4) suggested that the cells differentiated with our protocol formed stratified EEC layers but not piled-up collective cells without cell-cell adhesion. Notably, CK13 is expressed in the suprabasal and apical region but not in the basal region in authentic esophageal epithelium [8,21]; all layers were positive for CK13 in our hiPSC-derived tissues, indicating that the tissues did not identically recapitulate an authentic human esophageal epithelium structure.

Enhancing effects of ATRA on esophageal epithelial differentiation
We next evaluated the effects of ATRA in our esophageal differentiation protocol (Fig. 3a), as the effects of ATRA in the differentiation of FG have not been clarified [8,10,20]. First, we differentiated the cells into an esophageal lineage with 0, 0.1, 0.5 and 1.0 lM of ATRA from Day 6 (FG) to Day 21 (EEC) (S- Fig. 4a) to identify the appropriate concentration of ATRA for this study. Quantitative PCR showed that 0.5 and 1.0 lM ATRA treatment induced a higher expression of SOX2, CK13 and FABP5 than other concentrations (S- Fig. 4b). Considering the toxicity, the lower concentration of 0.5 lM was used for the subsequent experiments. The differentiation protocol devoid of ATRA obviously resulted in reduced staining of glycogen with lugol staining (Fig. 3b) and CK13 according to immunohistology (Fig. 3c), suggesting that ATRA plays a critical role in differentiation into esophagus.
We then analyzed the transcriptome data of the cells derived with or without ATRA. The scatterplot (Fig. 3d) and Z-score heatmap (Fig. 3e) showed the upregulation of 13 of the 17 esophagus-specific genes that were extracted from publicly available RNA-seq data (see the ''Material and Methods'' section) as well as known retinoid target genes [28] in the ATRA( ?) cells.
Furthermore, since RA signaling is reported to have positive effects on lung [20] and posterior FG differentiation [8], we examined the expression of marker genes of lung (SFTEB, SFTPC and NKX2.1) and stomach, which is the posterior FG derivative, (TFF1, TFF2 and MUC5AC) in ATRA( ?) differentiated cells using semi-quantitative RT-PCR (Fig. 3f, g). All of these marker genes were not expressed in the differentiated cells.
These data indicated that ATRA had a positive effect on the process of differentiation into esophageal epithelium.

Effects of ATRA on differentiation from FG to dAFG and from dAFG to EECs
The effects of ATRA on organ specification from FG have been controversial [8,20] (S- Fig. 5; also, see the ''discussion'' section). To clarify whether or not ATRA promotes differentiation of FG into dAFG in our protocol, we compared the expression of the dAFG marker genes SOX2 and p63 on Day 13 in the presence and absence of ATRA using quantitative RT-PCR (S- Fig. 6a). Our findings showed that ATRA significantly enhanced the expression of these marker genes as well as FABP5, a known target gene of ATRA (S- Fig. 6b).
We next assessed the effects of ATRA on the differentiation between dAFG and EEC (S- Fig. 6c), as no previous report has clarified the effect of ATRA treatment at this step (S- Fig. 5b, 5c). The results showed that, during transition from dAFG to EEC as well as from FG to dAFG, ATRA enhanced the expression of the EEC marker genes FABP5 and CK13 at Day 24 in the protocol (S- Fig. 6d).

Effects of ATRA on differentiation into EECs via RARc
ATRA upregulates the downstream genes through RAR and RXR heterodimers. Of the three subtypes of RARs, RARa and RARb are widely expressed in many tissues, while the expression of RARc is limited but particularly strong in the skin and esophagus (S- Fig. 7a) [18].
To identify which subtype of RARs is important for esophageal differentiation, we used the small molecules AM580 (RARa-specific agonist), CD2314 (RARb-specific agonist) and BMS961 (RARc-specific agonist) instead of ATRA ( Fig. 4a and S-Fig. 7b) from Day 6 (FG) to Day 21 (EEC). RT-PCR showed that AM580 and CD2314 treatment resulted in a non-significant increasing trend in the expression of FABP5, which is a retinoid target gene [28] and whose expression is highest in the esophagus among human organ tissues [18], and SOX2, a general FG marker, but not the EEC-specific markers CK13, PAX9 and S100A14 (Fig. 4b and S-Fig. 7b). In contrast, BMS961 treatment significantly enhanced the expression of FABP5, CK13 and S100A14, similarly to ATRA treatment (Fig. 4c).
Therefore, in order to confirm whether or not ATRA enhances the differentiation of human esophagus via RARc, we compared the expression of the esophageal markers in the resultant cells in the presence or absence of J Gastroenterol (2020) 55:763-774 771 the RARc antagonist LY2955303 during differentiation culture with ATRA (Fig. 4d). As shown in the Fig. 4e, RARc antagonist treatment cancelled the enhancing effects of ATRA on the expression of esophageal-related genes, such as FABP5, SOX2, CK13, PAX9 and S100A14. Furthermore, to exclude the possibility that some off-target effect of LY2955303 contributed to the results of the experiments, we also used another RARc antagonist, MM11253, and obtained the same results (S- Fig. 7c). In addition, we found that RARc was expressed in the FG (Day 6), and the expression increased as the differentiation progressed (Fig. 4f). These results indicate that the enhancing effect of ATRA on esophageal differentiation depends on RARc activation. In addition, to address whether or not the increased SOX2 expression induced by RARa and RARb agonist ( Fig. 4b and S-Fig. 7b) meant that these agonists induced differentiation into other tissues expressing SOX2, such as stomach and neuronal cells [18], we examined the expression of marker genes of stomach (TFF1, TFF2 and MUC5AC) and neuronal cells (SOX1 and PAX6) in the cells treated with RARa and RARb agonist by semiquantitative RT-PCR. None of these marker genes was expressed in the cells (S- Fig. 8a and b), indicating that neither RARa nor RARb agonist induced differentiation into these kinds of cells.
Treatment with an RARc-specific agonist instead of ATRA to induce stratified layers of cells expressing esophageal epithelial markers Finally, we examined whether or not treatment with an RARc agonist instead of ATRA could also induce cells that recapitulated the expression of esophageal differentiation marker proteins and stratified squamous epithelial structures (Fig. 5a). After culture with the RARc-specific agonist using ALI, immunohistological analyses showed stratified cells with the obvious expression of SOX2, p63, CK5, E-cadherin, CK13, PAX9, Claudin-4 and Involucrin ( Fig. 5b and S-Fig. 9).

Discussion
We successfully generated stratified cells expressing esophageal epithelial markers from hiPSCs with our original method and evaluated the significance of RA signaling on EEC induction after FG differentiation. RA signaling is known to be involved in the development of various tissues, such as brain, limbs, eye, lung and pancreas [29,30]. However, the effect of RA signaling on the esophageal development remains to be clarified.
Several previous studies using mouse models [13,14], but none using human cells, have reported that ATRA enhanced the terminal differentiation and proliferation of esophageal progenitor cells. However, these mouse studies did not address the role of RA signaling in the early steps of esophageal development. There have been only two previous papers on hiPSCs-derived esophageal differentiation. One of them did not use any RA signaling agonists at all in their protocol. The other (Triso et al. [8]) evaluated the effect of ATRA before (but not after) FG differentiation. The present study demonstrated that ATRA promotes both progression from FG to dAFG, which is an esophageal progenitor (S- Fig. 6b), and from dAFG to esophageal epithelial differentiation (S- Fig. 6d) using hiPSC-derived models. Thus, to our knowledge, this is the first ever report describing the continuous effects of RA signaling on the processes of human EEC differentiation from FG.
In the present study, we demonstrated the enhancing effect of continuous activation of RA signaling on EEC induction after FG differentiation. In previous reports, the effect of RA signaling on the induction of the derivatives of FG has been controversial. We drew a comparative diagram (S- Fig. 5) showing our current protocol (a), the previous protocol for differentiation into posterior FG (PFG) (b) and ventral anterior FG (vAFG) (c) from foregut. Triso et al. [8] reported that ATRA treatment for 1 day during FG induction from DE at Day 5 in their protocol resulted in the increased expression of p63 and GATA4 at Day 9. p63 is expressed in not only dAFG but vAFG as well, which is a progenitor of lung cells [20,31]. GATA4 is a marker of posterior FG (PFG) [8,18,32] but not AFG. The authors also reported that ATRA treatment for four days after Day 5 promoted posterior FG differentiation rather than AFG [8] (S-Fig 5b). Thus, which fate specification of FG into dAFG, vAFG and PFG was promoted by ATRA has been unclear. Green et al. [20] showed that six days' treatment of FG with ATRA and CKFBE upregulated NKX2.1, a lung marker (S- Fig 5c). However, they did not assess the expression of the AFG marker SOX2, so whether ATRA treatment promotes vAFG-specific differentiation or nonspecific AFG differentiation from FG has been unclear. In the present study, we demonstrated that treatment of FG with ATRA from Day 6 to Day 13 resulted in the enhanced expression of SOX2 and p63 (S- Fig. 6b), with no expression of NKX2.1 (S- Fig. 2a) or GATA4 (S- Fig. 2b). Furthermore, in our protocol, continuous treatment of FG with ATRA and CKFBE for 15-18 days enhanced neither lung markers, such as NKX2.1, SFTPB and SFTPC (Fig. 3f), nor PFG-derived stomach markers, such as TFF1, TFF2 and MUC5AC (Fig. 3g). These data indicated that ATRA promotes dAFG specification from FG. In addition, while Triso et al. [8] suggested that ATRA should be added prior to Day 6 (FG), we achieved efficient FG differentiation without ATRA by optimizing the concentration of CHIR during differentiation into DE (S- Fig. 1c and d). Differences in the identity of cells receiving ATRA among these reports might have led to these inconsistent outcomes of ATRA treatment.
No previous studies arguing the importance of retinoic acid in esophageal epithelial differentiation clarified which subtype of RA receptors played significant roles in the effect of RA. We identified for the first time RARc as the RAR subtype whose activation promotes human EEC differentiation. This is compatible with the fact that RARc is the predominant subtype of RAR in esophagus [18]. Regarding the differentiation of skin, which highly express RARc (S- Fig. 3a), previous studies using mouse models reported that RA signaling via RARc played an important role [28,33]. In the present study, RARa and RARb agonist treatment resulted in an increased expression of SOX2 without statistical significance, while no marked differences were noted in other esophageal markers ( Fig. 4b and S-Fig. 7b). SOX2 is an early marker in the process of differentiation toward esophagus [21], and its expression is reduced in mature esophageal cells. RARa and RARb agonist might promote the early phase of this process but not the maturation of esophageal cells. Notably, the combination of ATRA and RARc antagonist did not increase the SOX2 expression, whereas that of RARa and RARb agonist did. A previous study reported that ATRA binds to three types of RAR with different affinities [34]. ATRA may, therefore, have activated neither RARa nor RARb in the current experiments, while the agonists AM580 and CD2314 activated RARa and RARb, respectively, and enhanced the expression of SOX2.
Of note, several reports argued that RA signaling had a suppressive effect on esophageal cancer [35] as well as skin cancer [36]. In addition, the decreased expression of DSG3 [37], SPRR3 [38] and MAL [39], which were all also decreased under conditions of ATRA absence in our experiment (Fig. 3e), were reported to be related to a poor prognosis of esophageal cancer. However, no report has clarified the relationship between the prognosis and expression or mutation of RARc in esophageal cancer. In this context, we might need to focus on other hormones affecting esophageal differentiation as well. BMP4 and EGF were reported to promote esophageal differentiation from iPSCs [7] and proliferation of esophageal keratinocytes [40], respectively, so in the final step of our differentiation protocol, we employed BMP4 and EGF as well as ATRA and omitted the reagents (CHIR ? FGF10 ? KGF) reported to be required for inducing differentiation into the lung [20] in order to achieve specified differentiation to the esophagus. Future studies focusing on RARc signaling as well as these other relevant signals may provide some insight into not only esophageal differentiation but also the pathogenesis of esophageal cancer.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.